Research on State of Charge estimation method for an improved lithium-ion gas-liquid dynamics model with temperature parameters

Accurate estimation of the State of Charge (SOC) is fundamental to ensure the safe and reliable operation of lithium-ion batteries, and developing a high-performance battery model is crucial for achieving precise SOC estimation. However, current lithium-ion battery models struggle to balance accuracy and complexity. Against this background, on the basis of further investigating battery polarization phenomena and taking temperature effects into account, this paper constructs an improved gas–liquid dynamics model with temperature parameters and proposes an online SOC estimation method based on this model. The model offers good accuracy, and its voltage equation is only first-order, resulting in low computational cost. This model can more perfectly simulate the voltage rebound characteristics exhibited by the battery as a result of polarization reactions and takes into account the influence of temperature on SOC estimation. Compared with the original gas–liquid dynamics model, this enhanced model improves SOC estimation accuracy. Additionally, the unscented Kalman filter algorithm is introduced to achieve online SOC estimation. Finally, a 3Ah ternary lithium battery is tested under three driving cycles to validate the proposed model and algorithm. The results show that the maximum SOC estimation error is less than 2% in all cases. When facing initial SOC errors and initial temperature errors, the method rapidly eliminates the impact of initial errors, and the influence of initial errors on subsequent SOC estimation results is minimal, demonstrating strong robustness of the proposed algorithm. • Propose an IGLD-TP model that achieves higher accuracy than the original GLD model. • The IGLD-TP model with the UKF algorithm, achieves online SOC estimation. • The algorithm can quickly eliminate the influence of the initial SOC error. • The algorithm rapidly mitigates the impact of initial temperature error on SOC. • The algorithm has strong robustness against sampling noise.